PhD-Forum en Ingeniería Informática

Título de la charla: "From FANTASIA to PROTEA: Engineering a Computational Method into a Research Platform"

Resumen: PROTEA annotates protein function from protein language model embeddings. It did not start from scratch: it grew out of FANTASIA, an earlier embedding-based annotation tool. That lineage is the heart of this talk. Moving from a working research tool to a system that can scale, evolve, and be reproduced is, above all, a software engineering effort. We will see how a modular architecture split across independent repositories, with explicit contracts and a plugin system, makes it possible to swap the embedding model, the annotation source, or the experiment engine without rewriting the core. We will show how large-scale processing relies on a distributed system of queues and workers that tolerates failures and resumes jobs. We will explain why reproducibility is designed rather than promised: every artifact is identified by fingerprints of its schema and manifest, so an experiment can be repeated and audited. We will close with the automation that holds it all together (continuous integration, branch protection, and orchestration of the development work itself). The goal is to show, through a real case, what software engineering brings when a research method must scale, evolve, and outlive its authors.

Ponente: Francisco Miguel Pérez Canales.

Enlace al proyecto.